Estimating the Variance of the Maximum Pseudo-Likelihood Estimator
Lynne Seymour · Lecture notes-monograph series · 2001
The use of the pseudo-likelihood estimator for Gibbs-Markov random field models has a distinct advantage over more conventional approaches mainly due to its computational efficiency.Indeed, the maximum pseudo-likelihood estimator (MPLE) is often used as the Monte Carlo parameter in Markov chain Monte Carlo (MCMC) simulations.The MPLE itself has some very nice estimation properties, though its variance is still undiscovered.In this paper, the moving-block bootstrap is employed to estimate the variance of the MPLE in the Ising model.